Plotting using R

1. Week 4

  • Anatomy of ggplot
  • Scatter plots
  • Line plots

2. Week 5

  • Bar Plots
  • Histograms
  • Multiple geoms, multiple aes()
  • All other plots

3. Week 6, Tuning plots

  • Scales
  • Colors
  • Titles and labels
  • Themes
  • Save your plot

Sync your Repo

  • Open GitHub Desktop
  • Click “Current Branch” (should be “Main”)
  • You will see a branch of “upstream/main”. Select branch “Main”, click “Choose a branch to Merge into Main”.
  • Choose “upstream/main”, merge into your own “Main”. After merging, click “Push Origin”.

Get the data

We are using the gapminder dataset (https://www.gapminder.org/data) that has been put into an R package by @bryan2017 so we can load it with library(gapminder).

library(tidyverse)
library(gapminder)

Let’s create a new shorter tibble called gapdata2007 that only includes data for the year 2007.

gapdata2007 <- gapminder %>% 
  filter(year == 2007)

loads the gapminder dataset from the package environment into your Global Environment

gapdata <- gapminder

Both gapdata and gapdata2007 now show up in the Environment tab and can be clicked on/quickly viewed as usual.

Homework week 5

Create the following figure of life expectancy in American countries (year 2007): (if the figure seems too small in RStudio, open the ‘fig_week5_homework.png’ file in your homework folder)

fig

Hints:

  • If geom_bar() doesn’t work try geom_col() or vice versa.
  • coord_flip() to make the bars horizontal (it flips the x and y axes).
  • x = country gets the country bars plotted in alphabetical order, use x = fct_reorder(country, lifeExp) still inside the aes() to order the bars by their lifeExp values. Or try one of the other variables (pop, gdpPercap) as the second argument to fct_reorder().
  • when using fill = NA, you also need to include a color; we’re using color = "deepskyblue" inside the geom_col().
  • use geom_text() to label the lifeExp. You can use round() to round up to 1 digit.
  • Choose theme_classic()
gapdata2007 %>% 
  filter(continent == 'Americas') %>% 
  ggplot(aes(x = fct_reorder(country, lifeExp), y = lifeExp)) + 
  geom_col(fill = NA, color = "deepskyblue") + 
  coord_flip() + 
  geom_text(aes(label = round(lifeExp, digits = 1))) + 
  theme_classic()

Fine tuning plots

We can save a ggplot() object into a variable (we usually call it p but it can be any name). This then appears in the Environment tab. To plot it it needs to be recalled on a separate line to get drawn. Saving a plot into a variable allows us to modify it later (e.g., p + theme_bw()).

p0 <-  gapdata2007 %>% 
  ggplot(aes(y = lifeExp, x = gdpPercap, color = continent)) +
  geom_point(alpha = 0.3) +
  theme_bw() +
  geom_smooth(method = "lm", se = FALSE) +
  scale_color_brewer(palette = "Set1")

p0
p0: Starting plot for the examples in this chapter.

p0: Starting plot for the examples in this chapter.

Scales

Logarithmic

Transforming an axis to a logarithmic scale can be done by adding on scale_x_log10():

p1 <- p0 + scale_x_log10()
p1

scale_x_log10() and scale_y_log10() are shortcuts for the base-10 logarithmic transformation of an axis. The same could be achieved by using, e.g., scale_x_continuous(trans = "log10"). The latter can take a selection of options, namely "reverse", "log2", or "sqrt". Check the Help tab for scale_continuous() or look up its online documentation for a full list.

Expand limits

A quick way to expand the limits of your plot is to specify the value you want to be included:

p2 <- p0 + expand_limits(y = 0)
p2

Or two values for extending to both sides of the plot:

p3 <- p0 + expand_limits(y = c(0, 100))
p3

By default, ggplot() adds some padding around the included area (see how the scale doesn’t start from 0, but slightly before). This ensures points on the edges don’t get overlapped with the axes, but in some cases - especially if you’ve already expanded the scale, you might want to remove this extra padding. You can remove this padding with the expand argument:

p4 <- p0 +
  expand_limits(y = c(0, 100)) +
  coord_cartesian(expand = FALSE)

p4

We are now using a new library - patchwork - to print all 4 plots together. Its syntax is very simple - it allows us to add ggplot objects together. (Trying to do p1 + p2 without loading the patchwork package will not work, R will say “Error: Don’t know how to add p2 to a plot”.)

library(patchwork)
p1 + p2 + p3 + p4 + plot_annotation(tag_levels = "1", tag_prefix = "p")
p1: Using a logarithmic scale for the x axis. p2: Expanding the limits of the y axis to include 0. p3: Expanding the limits of the y axis to include 0 and 100. p4: Removing extra padding around the limits.

p1: Using a logarithmic scale for the x axis. p2: Expanding the limits of the y axis to include 0. p3: Expanding the limits of the y axis to include 0 and 100. p4: Removing extra padding around the limits.

Zoom in

p5 <- p0 +
  coord_cartesian(ylim = c(70, 85), xlim = c(20000, 40000)) 

p5

How is this one different to the previous?

p6 <- p0 +
  scale_y_continuous(limits = c(70, 85)) +
  scale_x_continuous(limits = c(20000, 40000))
  #ylim(70, 85) + xlim(20000,40000)

p6

p5 + labs(tag = "p5") + p6 + labs(tag = "p6")
p5: Using `coord_cartesian()` vs p6: Using `scale_x_continuous()` and `scale_y_continuous()` for setting the limits of plot axes.

p5: Using coord_cartesian() vs p6: Using scale_x_continuous() and scale_y_continuous() for setting the limits of plot axes.

Previously we used patchwork’s plot_annotation() function to create our multiplot tags. Since our exmaples no longer start the count from 1, we’re using ggplot()’s tags instead, e.g., labs(tag = "p5"). The labs() function will be covered in more detail later in this chapter.

Axis ticks

ggplot() does a good job deciding how many and which values include on the axis.

But sometimes you’ll want to specify these. We can do so by using the breaks argument:

# calculating the maximum value to be included in the axis breaks:
max_value <- gapdata2007$lifeExp %>% max() %>% round(1)

# using scale_y_continuous(breaks = ...):
p7 <-  p0 +
  coord_cartesian(ylim = c(0, 100), expand = FALSE) +
  scale_y_continuous(breaks = c(18, 50, max_value))

# we may also include custom labels for our breaks:
p8 <-  p0 +
  coord_cartesian(ylim = c(0, 100), expand = FALSE) +
  scale_y_continuous(breaks = c(18, 50, max_value), labels = c("Adults", "50", "MAX"))

p7 + labs(tag = "p7") + p8 + labs(tag = "p8")
p7: Specifiying y axis breaks. p8: Adding custom labels for our breaks.

p7: Specifiying y axis breaks. p8: Adding custom labels for our breaks.

Colors

Using the Brewer palettes:

The easiest way to change the color palette of your ggplot() is to specify a Brewer palette.

p9 <- p0 +
  scale_color_brewer(palette = "Paired")
p9

Note that http://colorbrewer2.org/ also has options for Colorblind safe and Print friendly.

Another easy way for scientific publications is to use package ggsci.

library(ggsci)

p9 <- p0 +
  scale_color_jama()
p9

Legend title

scale_color_brewer() or scale_color_jama() is also a convenient place to change the legend title.

library(ggsci)

p10 <- p0 +
  scale_color_rickandmorty(name = "Continent - \n one of 5")
p10

Note the \n inside the new legend title - new line.

p9 + labs(tag = "p9") + p10 + labs(tag = "p10")
p9: Choosing a ggsci palette for your colors. p10: Changing the legend title.

p9: Choosing a ggsci palette for your colors. p10: Changing the legend title.

Choosing colors manually

R also knows the names of many colors, so we can use words to specify colors:

p11 <- p0 +
  scale_color_manual(values = c("red", "green", "blue", "purple", "pink"))
p11

The same function can also be used to use HEX codes for specifying colors:

p12 <- p0 +
  scale_color_manual(values = c("#8dd3c7", "#ffffb3", "#bebada",
                                "#fb8072", "#80b1d3"))
p12

p11 + labs(tag = "p11") + p12 + labs(tag = "p12")
Colors can also be specified using words (`"red"`, `"green"`, etc.), or HEX codes (`"#8dd3c7"`, `"#ffffb3"`, etc.).

Colors can also be specified using words ("red", "green", etc.), or HEX codes ("#8dd3c7", "#ffffb3", etc.).

Titles and labels

We’ve been using the labs(tag = ) function to add tags to plots. But the labs() function can also be used to modify axis labels, or to add a title, subtitle, or a caption to your plot.

p13 <- p0 +
  labs(x = "Gross domestic product per capita",
       y = "Life expectancy",
       title = "Health and economics",
       subtitle = "Gapminder dataset, 2007",
       caption = Sys.Date() %>% format("%B %d, %Y"),
       tag = "p13")

p13
p13: Adding on a title, subtitle, caption using `labs()`.

p13: Adding on a title, subtitle, caption using labs().

Annotation

In the previous chapter, we showed how use geom_text() and geom_label() to add text elements to a plot. Using geoms make sense when the values are based on data and variables mapped in aes(). They are efficient for including multiple pieces of text or labels on your data. For ‘hand’ annotating a plot, the annotate() function makes more sense, as you can quickly insert the type, location and label of your annotation.

p14 <- p0 +
  annotate("text",
           x = 25000,
           y = 50,
           label = "No points here!")
p15 <- p0 +
  annotate("label",
           x = 25000,
           y = 50,
           label = "No points here!")
p16 <- p0 +
  annotate("label",
           x = 25000, 
           y = 50,
           label = "No points here!", 
           hjust = 0)
p14 + labs(tag = "p14") + (p15 + labs(tag = "p15"))/ (p16 + labs(tag = "p16"))
p14: `annotate("text", ...)` to quickly add a text on your plot. p15: `annotate("label")` is similar but draws a box around your text (making it a label). p16: Using `hjust` to control the horizontal justification of the annotation.

p14: annotate("text", ...) to quickly add a text on your plot. p15: annotate("label") is similar but draws a box around your text (making it a label). p16: Using hjust to control the horizontal justification of the annotation.

hjust stands for horizontal justification. Its default value is 0.5 (see how the label was centered at 25,000 - our chosen x location), 0 means the label goes to the right from 25,000, 1 would make it end at 25,000.

Annotation with a superscript and a variable

This is an advanced example on how to annotate your plot with something that has a superscipt and is based on a single value read in from a variable.

# a value we made up for this example
# a real analysis would get it from the linear model object
fit_glance <- tibble(r.squared = 0.7693465)


plot_rsquared <- paste0(
  "R^2 == ",
  fit_glance$r.squared %>% round(2))


p17 <- p0 +
  annotate("text",
           x = 25000, 
           y = 50,
           label = plot_rsquared, parse = TRUE,
           hjust = 0)

p17 + labs(tag = "p17")
p17: Using a superscript in your plot annotation.

p17: Using a superscript in your plot annotation.

Overall look

And finally, everything else on a plot - from font to background to the space between facets, can be changed using the theme() function. As you saw in the previous chapter, in addition to its default grey background, ggplot2 also comes with a few built-in themes, namely, theme_bw() or theme_classic(). These produce good looking plots that may already be publication ready. But if we do decide to tweak them, then the main theme() arguments we use are axis.text, axis.title, and legend.position.

Note that all of these go inside the theme(), and that the axis.text and axis.title arguments are usually followed by = element_text() as shown in the examples below.

Text size

The way the axis.text and axis.title arguments of theme() work is that if you specify .x or .y it gets applied on that axis alone. But not specifying these, applies the change on both. Both the angle and vjust (vertical justification) options can be useful if your axis text doesn’t fit well and overlaps. It doesn’t usually make sense to change the color of the font to anything other than "black", we are using green and red here to indicate which parts of the plot get changed with each line.

p18 <-  p0 +
  theme(axis.text.y = element_text(color = "green", size = 14),
        axis.text.x = element_text(color = "red",  angle = 45, vjust = 0.5, face = 'bold'),
        axis.title  = element_text(color = "blue", size = 16, family = "Times", face = 'italic')
        )

p18 + labs(tag = "p18")
p18: Using `axis.text` and `axis.title` within `theme()` to tweak the appearance of your plot, including font size and angle. Colored font is used to indicate which part of the code was used to change each element.

p18: Using axis.text and axis.title within theme() to tweak the appearance of your plot, including font size and angle. Colored font is used to indicate which part of the code was used to change each element.

Legend position

The position of the legend can be changed using the legend.position argument within theme(). It can be positioned using the following words: "right", "left", "top", "bottom". Or to remove the legend completely, use "none":

p19 <- p0 +
  theme(legend.position = "none")
p19

Alternatively, we can use relative coordinates (0–1) to give the legend a relative x-y location.

p20 <- p0 +
  theme(legend.position      = c(1,0), #bottom-right corner
        legend.justification = c(1,0)) 
p20

p19 + labs(tag = "p19") + p20 + labs(tag = "p20")
p19: Setting `theme(legend.position = "none")` removes it. p20: Relative coordinates such as `theme(legend.position = c(1,0)` can by used to place the legend within the plot area.

p19: Setting theme(legend.position = "none") removes it. p20: Relative coordinates such as theme(legend.position = c(1,0) can by used to place the legend within the plot area.

Further theme(legend.) options can be used to change the size, background, spacing, etc., of the legend. However, for modifying the content of the legend, you’ll have to use the guides() function. Again, ggplot()’s defaults are very good, and we rarely need to go into this much tweaking using both the theme() and guides() functions. But it is good to know what is possible.

For example, this is how to change the number of columns within the legend (Figure @ref(fig:chap05-fig-p21)):

p21 <- p0 +
  guides(color = guide_legend(ncol = 2)) +
  theme(legend.position = "top") # moving to the top optional

p21 + labs(tag = "p21")
p21: Changing the number of columns within a legend.

p21: Changing the number of columns within a legend.

Saving your plot

There are several ways to save your figures.

  • Save in Rstudio at each code chunk, right click the images. (NOT RECOMMENDED)
  • Knit to an html, pdf or word document, and save the images from those files.
  • Or, use ggsave() function in your code. It can save a single plot into a variety of formats, namely "pdf" or "png"
ggsave(p0, file = "my_saved_plot.pdf", width = 5, height = 4)

If you omit the first argument - the plot object - and call, e.g., ggsave(file = "plot.png) it will just save the last plot that got printed.

Text size tip: playing around with the width and height options (they’re in inches) can be a convenient way to increase or decrease the relative size of the text on the plot. Look at the relative font sizes of the two versions of the ggsave() call, one 5x4, the other one 10x8.

ggsave(p0, file = "my_saved_plot_larger.pdf", width = 10, height = 8)

Again, we emphasis the importance of understanding the underlying data through visualization, rather than relying on statistical tests or, heaven forbid, the p-value alone.